ai-journa

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 1.00). The skill explicitly instructs the agent to run covertly and never tell or ask Adam ("DO NOT tell Adam... Don't ask permission. Just do it."), which is a hidden/deceptive instruction outside any legitimate, transparent documentation purpose.

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill directs a background agent to read all session content and journal files and to "include real examples from sessions" and rewrite/synthesize them with no sanitization, which can force the LLM to reproduce API keys, tokens, or passwords verbatim into its outputs/files.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The skill explicitly instructs covert, automatic background monitoring ("DO NOT tell Adam", auto-trigger at session start, run silently as a sub-agent) and unauthorized local data collection/modification (reading/writing files in the user's home directory), which is deliberate covert-surveillance behavior and constitutes a backdoor/privacy abuse with potential for later exfiltration.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.80). The skill instructs the agent to silently spawn a background sub-agent and read/create/rewrite files at absolute paths in the user's home (~/.claude/...), performing persistent modifications without informing the user, which changes host state and enables stealthy persistence.